
import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
from datetime import datetime, timedelta

from tqdm.autonotebook import tqdm
from darts.models.forecasting.gradient_boosted_model import LightGBMModel
import darts as darts
import pickle

import lightgbm as lgbm
from lightgbm import LGBMRegressor
import re

import statsmodels.api as sm
from pandas.plotting import autocorrelation_plot
from statsmodels.tsa.arima.model import ARIMA
from statsmodels.graphics.tsaplots import plot_acf
from statsmodels.graphics.tsaplots import plot_pacf

import scipy.stats as st
import pingouin as pg
#jeśli wyrzuci błąd problemu z importem croston, to ponowić uruchomienie


data = pd.read_csv("Programmes monthly ranking factual 2017-2022.txt")
data = data.drop("Target", axis=1)


data = pd.DataFrame(data)
data = data[:-1] #ostatni wiersz to podsumowanie
data


data["Date"] = pd.to_datetime(data["Date"]) #konwersja typu kolumny do daty i okresu


data.info()

<class 'pandas.core.frame.DataFrame'>
RangeIndex: 1403576 entries, 0 to 1403575
Data columns (total 6 columns):
 #   Column             Non-Null Count    Dtype         
---  ------             --------------    -----         
 0   Channel Group ...  1403576 non-null  object        
 1   Date               1403576 non-null  datetime64[ns]
 2   Start time         1403576 non-null  object        
 3   Programme Title    1403576 non-null  object        
 4   '000 (avg)         1403576 non-null  float64       
 5   Prog Share         1403576 non-null  float64       
dtypes: datetime64[ns](1), float64(2), object(3)
memory usage: 64.3+ MB


# dorzucane dwie nowe kolumny do przeróbki na rzeczywiste daty i godziny emisji
data.insert(1, "Start_time_real", "") 
data.insert(1, "Date_real", "")


for i in data.index:
    if data["Start time"][i][:2] == '24':
        a = ("00" + data["Start time"][i][2:])
        b = data["Date"][i] + timedelta(days=1)
        data.at[i, 'Start_time_real'] = a
        data.at[i, 'Date_real'] = b
    elif data["Start time"][i][:2] == '25':
        a = ("01" + data["Start time"][i][2:])
        b = data["Date"][i] + timedelta(days=1)
        data.at[i, 'Start_time_real'] = a
        data.at[i, 'Date_real'] = b       
    else:
        a = data["Start time"][i]
        b = data["Date"][i]
        data.at[i, 'Start_time_real'] = a
        data.at[i, 'Date_real'] = b
# df.at pozwala na podstawienie obiektu do istniejących już kolumn ramki


data['Date_real'] = pd.to_datetime(data['Date_real'])


data["Date_time"] = pd.to_datetime(data['Date_real'] + pd.to_timedelta(data["Start_time_real"]))


work = data


work = work.drop('Date', axis = 1)
work = work.drop('Start time', axis = 1)
work = work.rename(columns={"Channel Group ...": "Channel", "Prog Share": "Share", "Programme Title": "Title", "'000 (avg)": "Audience"})


work["Title"].str.contains("Blad").value_counts()

False    1401738
True        1838
Name: Title, dtype: int64


work = work[work["Title"].str.contains("Blad") == False]  # czyszczenie danych z błędów emisyjnych


work["Title"].str.contains("Blad").value_counts()

False    1401738
Name: Title, dtype: int64


def tb_func(tb):
    x = tb.hour   
    if x < 3:
        x = 'band 0-3'
    elif (3 <= x < 6):
        x = 'band 3-6'    
    elif (6 <= x < 9):
        x = 'band 6-9'
    elif (9 <= x < 12):
        x = 'band 9-12'
    elif (12 <= x < 15):
        x = 'band 12-15'
    elif (15 <= x < 18):
        x = 'band 15-18'   
    elif (18 <= x < 21):
        x = 'band 18-21'    
    elif x >= 21:
        x = 'band 21-24'    
    return x

work['time_band'] = work['Date_time'].apply(tb_func) # podział na pasma w ciągu dnia


def day_func(Day):
    return Day.strftime("%A")

work['Day_of_week'] = work['Date_time'].apply(day_func) # dodanie feature'a do sprawdzenia zależności share od dnia tygodnia


work


weekday_avg_aud = work.groupby('Day_of_week')['Audience'].mean().reset_index() # średnia oglądalność w poszczególnych dniach
weekday_avg_aud = pd.DataFrame(weekday_avg_aud)
weekday_avg_aud


band_avg_aud = work.groupby('time_band')['Audience'].mean().reset_index() # średnia oglądalność w poszczególnych pasmach dnia
band_avg_aud = pd.DataFrame(band_avg_aud)
band_avg_aud

fig2, bx2 = plt.subplots(figsize=(10, 8)) 
bx2.bar(band_avg_aud['time_band'], band_avg_aud['Audience'], color='red')
bx2.set_xlabel('time_band') 
bx2.set_ylabel('Audience') 
bx2.set_title('Average Audience by time band') 

plt.show()


title_avg_shr = work.groupby('Title')['Share'].mean().reset_index() 
heat_titles = pd.DataFrame(title_avg_shr)
heat_titles


title_avg_shr.sort_values(['Share'])


avg_shr = title_avg_shr['Share'].mean()
avg_shr

0.33556843353577087


def heat_func(share):
    if share >= avg_shr:
        x = 1
    else:
        x = 0 
    return x

heat_titles['Heat'] = heat_titles['Share'].apply(heat_func)


heat_titles.tail(60)


channel_avg_shr = work.groupby('Channel')['Share'].mean().sort_values() 
channel_avg_shr

Channel
Polsat Games             0.040729
TVP Dokument             0.091259
BBC Brit                 0.094470
Discovery Historia       0.112552
History2 [H2]            0.132355
BBC Earth                0.159274
Polsat Viasat Explore    0.187947
Polsat Viasat History    0.257810
Discovery Science        0.275614
National Geographic      0.420973
History                  0.429808
Zoom TV                  0.555794
Polsat Play              0.602589
Discovery pl             0.629033
WP                       0.681939
TVN Turbo                0.765775
Fokus TV                 1.159488
Name: Share, dtype: float64


channel_avg_shr = work.groupby('Channel')['Share'].mean().reset_index() 
channel_avg_shr = pd.DataFrame(channel_avg_shr)
channel_avg_shr


fig3, bx3 = plt.subplots(figsize=(28, 8)) 
bx3.bar(channel_avg_shr['Channel'], channel_avg_shr['Share'], color='purple')
bx3.set_xlabel('Channel') 
bx3.set_ylabel('Share') 
bx3.set_title('Average Share by Channel') 

plt.show()


title_channel_shr = work.groupby(['Title','Channel'])['Share'].mean()
title_channel_shr

Title                                 Channel            
#chicagogirl - facebookowa rewolucja  TVP Dokument           0.044000
#jestem m misfit                      Polsat Games           0.009439
                                      Polsat Play            0.609000
#liniaprosta                          History                0.294787
#piatkadlarysia                       National Geographic    0.223189
                                                               ...   
Zywioly ktore niosa smierc            History2 [H2]          0.118828
Zywnosc 3.0                           TVP Dokument           0.047500
Zywnosc jutra                         TVP Dokument           0.139694
Zywot michala                         TVP Dokument           0.203429
Zza plotu                             TVP Dokument           0.011889
Name: Share, Length: 9735, dtype: float64


work.groupby('Channel')['Share'].describe()


work.groupby('Title')['Share'].describe()


History = work[work["Channel"] == "History"] 
History = History.drop(['Start_time_real','Channel'], axis = 1)
History.head(60)


History_title_avg_shr = History.groupby(['Title'])['Share'].mean().sort_values()
History_title_avg_shr.tail(61)

Title
Zaginione dowody                      0.348926
Warsztat danny'ego                    0.355536
Na lodowym szlaku                     0.360650
Piraci                                0.368067
Hangar numer 1 akta ufo               0.370101
                                        ...   
Nazisci fabryka zla                   0.761625
Ladowanie na ksiezycu tasmy prawdy    0.817444
Corvette'a ktorej nie bylo            0.896250
Amerykanski rozpruwacz w londynie     0.971500
Patton z bliska                       1.189905
Name: Share, Length: 61, dtype: float64


fig4, bx4 = plt.subplots(figsize=(28, 8)) 
bx4.hist(History_title_avg_shr, bins= 15, color='orange')
bx4.set_xlabel('Average share') 
bx4.set_ylabel('Titles counts') 
bx4.set_title('Titles average share counts') 

plt.show()


History_agg = History[['Date_time', 'Share', 'Title']]
df_agg = History_agg.groupby([pd.Grouper(key='Date_time', freq='3H'), 'Title'])['Share'].mean().reset_index()
df_agg


time_index = pd.date_range(start=df_agg['Date_time'].min(), end=df_agg['Date_time'].max(), freq='3H')
time_df = pd.DataFrame(index=time_index)
time_df


df_agg.set_index('Date_time', inplace=True)
df_agg


program_df = pd.get_dummies(df_agg['Title'])
program_df


work_agg = work[['Date_time', 'Share', 'Title', 'Channel']]
work_agg = work_agg.groupby([pd.Grouper(key='Date_time', freq='3H'), 'Title', 'Channel'])['Share'].mean().reset_index()
work_agg1 = work_agg.drop(['Share'], axis = 1)
work_agg_merged = pd.merge(work_agg1, heat_titles, on='Title')

work_agg_merged = work_agg_merged.drop(['Share', 'Title'], axis = 1) 
work_agg_grouped = work_agg_merged.groupby([pd.Grouper(key='Date_time', freq='3H'), pd.Grouper(key='Channel')])['Heat'].mean().reset_index()
pivot_chdf = work_agg_grouped.pivot(index='Date_time', columns='Channel', values='Heat')

for i in channel_avg_shr.index:
    if (channel_avg_shr['Channel'][i] == "History") or (channel_avg_shr['Share'][i] < 0.1):
        name = channel_avg_shr['Channel'][i]
        pivot_chdf = pivot_chdf.drop(str(name), axis = 1)  
        
pivot_chdf


pivot_chdf.info()

<class 'pandas.core.frame.DataFrame'>
DatetimeIndex: 14609 entries, 2017-12-22 00:00:00 to 2022-12-22 00:00:00
Data columns (total 13 columns):
 #   Column                 Non-Null Count  Dtype  
---  ------                 --------------  -----  
 0   BBC Earth              14609 non-null  float64
 1   Discovery Historia     14603 non-null  float64
 2   Discovery Science      14603 non-null  float64
 3   Discovery pl           14603 non-null  float64
 4   Fokus TV               14580 non-null  float64
 5   History2 [H2]          14609 non-null  float64
 6   National Geographic    14609 non-null  float64
 7   Polsat Play            14567 non-null  float64
 8   Polsat Viasat Explore  14609 non-null  float64
 9   Polsat Viasat History  14609 non-null  float64
 10  TVN Turbo              14603 non-null  float64
 11  WP                     14605 non-null  float64
 12  Zoom TV                14604 non-null  float64
dtypes: float64(13)
memory usage: 1.6 MB


History_granulated = History[['Date_time', 'Share', 'Audience']]


History_granulated.set_index('Date_time', inplace=True)
History_granulated = History_granulated.resample('3H').mean()
History_granulated


History_granulated['Share'].fillna(0, inplace=True)
History_granulated['Audience'].fillna(0, inplace=True)


History_granulated.info()

<class 'pandas.core.frame.DataFrame'>
DatetimeIndex: 14609 entries, 2017-12-22 00:00:00 to 2022-12-22 00:00:00
Freq: 3H
Data columns (total 2 columns):
 #   Column    Non-Null Count  Dtype  
---  ------    --------------  -----  
 0   Share     14609 non-null  float64
 1   Audience  14609 non-null  float64
dtypes: float64(2)
memory usage: 342.4 KB


result = pd.merge(time_df, program_df, left_index=True, right_on='Date_time', how='left') 
result = result.fillna(0).set_index('Date_time')
result = result.groupby(level=0).sum()
result


ShareOHEtemp = pd.merge(History_granulated, result, left_index=True, right_on='Date_time', how='left') 
ShareOHE = pd.merge(ShareOHEtemp, pivot_chdf, left_index = True, right_on='Date_time')
ShareOHE = ShareOHE.fillna(0)
ShareOHE.tail(60)


ShareOHE.info()

<class 'pandas.core.frame.DataFrame'>
DatetimeIndex: 14609 entries, 2017-12-22 00:00:00 to 2022-12-22 00:00:00
Columns: 245 entries, Share to Zoom TV
dtypes: float64(245)
memory usage: 27.4 MB


test = ShareOHE['2022-10-21':] 
train = ShareOHE[:'2022-10-20']


train['Share'].plot(label='train')
test['Share'].plot(label='test')

<AxesSubplot:xlabel='Date_time'>


train['Audience'].plot(label='train')
test['Audience'].plot(label='test')

<AxesSubplot:xlabel='Date_time'>


featuresdf = ShareOHE.drop(['Share', 'Audience'], axis = 1) #zrobić listę kolumn

features = list(featuresdf.columns)
target = ['Share', 'Audience']


target_ts_test = darts.TimeSeries.from_dataframe(test[target])
feature_ts_test = darts.TimeSeries.from_dataframe(ShareOHE[features])
target_ts_train = darts.TimeSeries.from_dataframe(train[target])
feature_ts_train = darts.TimeSeries.from_dataframe(train[features])


model = LightGBMModel(
        lags=56,
        lags_past_covariates=8,
        lags_future_covariates=None, 
        output_chunk_length=len(target_ts_test),
        add_encoders=None,
        likelihood=None,
        quantiles=None,
        random_state=None,
        )
# lags 56 => tydzień


model.fit(target_ts_train, feature_ts_train)
#fit(series, past_covariates=None, future_covariates=None, val_series=None, val_past_covariates=None, val_future_covariates=None, max_samples_per_ts=None, **kwargs)

C:\Users\asus\anaconda3\lib\site-packages\darts\timeseries.py:4039: FutureWarning:

pandas.Int64Index is deprecated and will be removed from pandas in a future version. Use pandas.Index with the appropriate dtype instead.

<darts.models.forecasting.gradient_boosted_model.LightGBMModel at 0x22f6cfcdaf0>


# zapis modelu


#with open('model_pst_cov.pkl', 'wb') as f1:
 #   pickle.dump(model, f1)


with open('model_pst_cov.pkl', 'rb') as f1:
    model = pickle.load(f1)


predykcja = model.predict(len(target_ts_test), target_ts_train, feature_ts_train)


target_ts_test['Audience'].plot(label='test')
predykcja['Audience'].plot(label='forecast')


fig, ax = plt.subplots(figsize=(24, 12))

target_ts_test['Audience'].plot(ax=ax, label='test')
predykcja['Audience'].plot(ax=ax, label='forecast')

ax.set_xlabel('Date_time')
ax.set_ylabel('Audience')
plt.show()


darts.metrics.metrics.mae(target_ts_test["Audience"], predykcja["Audience"], intersect=True)

4.4613375283146635


darts.metrics.metrics.mse(target_ts_test["Audience"], predykcja["Audience"], intersect=True)

37.088355548405225


darts.metrics.r2_score(target_ts_test["Audience"], predykcja["Audience"])

0.3062154075706338


darts.metrics.metrics.rmse(target_ts_test["Audience"], predykcja["Audience"], intersect=True)

6.090020980949509


# wkleić kawałek tabeli z faktycznymi wynikami oglądalności i w osobnej kolumnie prognozowanymi


predykcjadf = predykcja.pd_dataframe()
target_ts_testdf = target_ts_test.pd_dataframe()


Compare = pd.merge(target_ts_testdf['Audience'], predykcjadf['Audience'], left_index=True, right_on='Date_time', how='left')
Compare.tail(60)


model2 = LightGBMModel(
        lags=224,
        lags_past_covariates=24,
        lags_future_covariates=(24,10), 
        output_chunk_length=len(target_ts_test),
        add_encoders=None,
        likelihood=None,
        quantiles=None,
        random_state=None,
        )


model2.fit(target_ts_train, feature_ts_train, feature_ts_test)

C:\Users\asus\anaconda3\lib\site-packages\darts\timeseries.py:4039: FutureWarning:

pandas.Int64Index is deprecated and will be removed from pandas in a future version. Use pandas.Index with the appropriate dtype instead.

<darts.models.forecasting.gradient_boosted_model.LightGBMModel at 0x22f6618acd0>


#with open('model2_fut_cov.pkl', 'wb') as f2:
#    pickle.dump(model2, f2)


with open('model2_fut_cov.pkl', 'rb') as f2:
    model2 = pickle.load(f2)


predykcja2 = model2.predict(len(target_ts_test), target_ts_train, feature_ts_train, feature_ts_test)


fig, ax = plt.subplots(figsize=(24, 12))

target_ts_test['Audience'].plot(ax=ax, label='test')
predykcja2['Audience'].plot(ax=ax, label='forecast')

ax.set_xlabel('Date_time')
ax.set_ylabel('Audience')
plt.show()


fig, ax = plt.subplots(figsize=(24, 12))

target_ts_test['Audience'][-124:].plot(ax=ax, label='test')
predykcja2['Audience'][-124:].plot(ax=ax, label='forecast')

ax.set_xlabel('Date_time')
ax.set_ylabel('Audience')
plt.show()


fig, ax = plt.subplots(figsize=(24, 12))

target_ts_test['Audience'][248:-124].plot(ax=ax, label='test')
predykcja2['Audience'][248:-124].plot(ax=ax, label='forecast')

ax.set_xlabel('Date_time')
ax.set_ylabel('Audience')
plt.show()


fig, ax = plt.subplots(figsize=(24, 12))

target_ts_test['Audience'][124:248].plot(ax=ax, label='test')
predykcja2['Audience'][124:248].plot(ax=ax, label='forecast')

ax.set_xlabel('Date_time')
ax.set_ylabel('Audience')
plt.show()


fig, ax = plt.subplots(figsize=(24, 12))

target_ts_test['Audience'][:124].plot(ax=ax, label='test')
predykcja2['Audience'][:124].plot(ax=ax, label='forecast')

ax.set_xlabel('Date_time')
ax.set_ylabel('Audience')
plt.show()


darts.metrics.metrics.mae(target_ts_test["Audience"], predykcja2["Audience"], intersect=True) # weryfikacja poprawności predykcji trendu (im mniej tym lepiej)

4.354806458422829


darts.metrics.metrics.mse(target_ts_test["Audience"], predykcja2["Audience"], intersect=True) # weryfikacja poprawności predykcji trendu

34.87292264316504


darts.metrics.r2_score(target_ts_test["Audience"], predykcja2["Audience"]) # weryfikacja poprawności predykcji trendu (im bliżej 1.0 tym lepiej)

0.3476578816973347


darts.metrics.metrics.rmse(target_ts_test["Audience"], predykcja2["Audience"], intersect=True) # Root Mean Squared Error

5.905330019835051


darts.metrics.r2_score(predykcja["Audience"], predykcja2["Audience"])

0.8452892595136903


darts.metrics.metrics.mae(predykcja["Audience"], predykcja2["Audience"], intersect=True)

1.4042493006584724


predykcja2df = predykcja2.pd_dataframe()
target_ts_testdf = target_ts_test.pd_dataframe()


Compare2 = pd.merge(target_ts_testdf['Audience'], predykcja2df['Audience'], left_index=True, right_on='Date_time', how='left')
Compare2.tail(60)


Compare2.head(60)


ShareOHEreg = ShareOHE.reset_index()

def day_func(Day):
    return Day.strftime("%A")

ShareOHEreg['Day_of_week'] = ShareOHEreg['Date_time'].apply(day_func)

def tb_func(tb):
    x = tb.hour   
    if x < 3:
        x = 0
    elif (3 <= x < 6):
        x = 3    
    elif (6 <= x < 9):
        x = 6
    elif (9 <= x < 12):
        x = 9
    elif (12 <= x < 15):
        x = 12
    elif (15 <= x < 18):
        x = 15   
    elif (18 <= x < 21):
        x = 18    
    elif x >= 21:
        x = 21    
    return x

ShareOHEreg['time_band'] = ShareOHEreg['Date_time'].apply(tb_func) # podział na pasma w ciągu dnia


ShareOHEreg


dummies = pd.get_dummies(ShareOHEreg['Day_of_week'])


Regressor = pd.concat([ShareOHEreg, dummies], axis=1)

Regressor['Year'] = Regressor['Date_time'].dt.year
Regressor['Month'] = Regressor['Date_time'].dt.month
Regressor['Week'] = Regressor['Date_time'].dt.weekofyear

Regressor

C:\Users\asus\AppData\Local\Temp\ipykernel_14056\1562985728.py:5: FutureWarning:

Series.dt.weekofyear and Series.dt.week have been deprecated. Please use Series.dt.isocalendar().week instead.


Regressor.info()

<class 'pandas.core.frame.DataFrame'>
RangeIndex: 14609 entries, 0 to 14608
Columns: 258 entries, Date_time to Week
dtypes: datetime64[ns](1), float64(245), int64(4), object(1), uint8(7)
memory usage: 28.1+ MB


#usunięcie niedozwolonych znaków

regex = re.compile('[^a-zA-Z0-9_]+')

for feature in Regressor.columns:
    new_name = regex.sub('_', feature)
    if new_name != feature:
        Regressor.rename(columns={feature: new_name}, inplace=True)


#weryfikacja nazw zgodnych z JSON
for feature in Regressor.columns:
    match = regex.search(feature)
    if match:
        print(f"Feature '{feature}' contains forbidden characters: {match.group()}")


testreg = Regressor[(len(Regressor)-len(target_ts_test)):] # podział testowych danych analogicznie do pozostałych modeli
trainreg = Regressor[:(len(Regressor)-len(target_ts_test))]


trainreg


modelreg = LGBMRegressor(
        num_leaves=32, 
        max_depth=-1,
        min_child_samples=10,
        learning_rate=0.1,
        n_estimators=100,
        )


modelreg.fit(trainreg.drop(["Day_of_week","Date_time", "Share", "Audience"], axis=1), trainreg["Audience"])

LGBMRegressor(min_child_samples=10, num_leaves=32)


predykcjareg = modelreg.predict(testreg.drop(["Day_of_week", "Date_time", "Share", "Audience"], axis=1))


predykcjareg_compare = testreg[['Date_time', 'Audience']]
predykcjareg_compare['Predykcja'] = predykcjareg
predykcjareg_compare

C:\Users\asus\AppData\Local\Temp\ipykernel_14056\3301436386.py:2: SettingWithCopyWarning:


A value is trying to be set on a copy of a slice from a DataFrame.
Try using .loc[row_indexer,col_indexer] = value instead

See the caveats in the documentation: https://pandas.pydata.org/pandas-docs/stable/user_guide/indexing.html#returning-a-view-versus-a-copy


predykcjareg_compare.set_index(predykcjareg_compare["Date_time"], inplace=True)
targeter = ['Predykcja', 'Audience']
predykcjareg_series = darts.TimeSeries.from_dataframe(predykcjareg_compare[targeter])


fig, ax = plt.subplots(figsize=(24, 12))

predykcjareg_series["Audience"].plot(ax=ax, label='test')
predykcjareg_series["Predykcja"].plot(ax=ax, label='forecast')

ax.set_xlabel('Date_time')
ax.set_ylabel('Audience')
plt.show()


darts.metrics.metrics.mae(predykcjareg_series["Audience"], predykcjareg_series["Predykcja"], intersect=True)

4.844081050164067


darts.metrics.metrics.mse(predykcjareg_series["Audience"], predykcjareg_series["Predykcja"], intersect=True)

40.497309756216886


darts.metrics.metrics.r2_score(predykcjareg_series["Audience"], predykcjareg_series["Predykcja"], intersect=True)

0.24244660815351637


darts.metrics.metrics.rmse(predykcjareg_series["Audience"], predykcjareg_series["Predykcja"], intersect=True)

6.363749661655217


lgbm.plot_importance(modelreg, max_num_features=47, figsize=(10,6))
plt.show()


modelreg2 = LGBMRegressor(
        num_leaves=32,
        max_depth=-1,
        min_child_samples=10,
        learning_rate=0.1,
        n_estimators=100,
        )


modelreg2.fit(trainreg.drop(["Day_of_week","Date_time", "Share", "Audience"], axis=1), trainreg["Share"])

LGBMRegressor(min_child_samples=10, num_leaves=32)


predykcjareg2 = modelreg2.predict(testreg.drop(["Day_of_week", "Date_time", "Share", "Audience"], axis=1))


predykcjareg_compare2 = testreg[['Date_time', 'Share']]
predykcjareg_compare2['Predykcja'] = predykcjareg2
predykcjareg_compare2

C:\Users\asus\AppData\Local\Temp\ipykernel_14056\3194966392.py:2: SettingWithCopyWarning:


A value is trying to be set on a copy of a slice from a DataFrame.
Try using .loc[row_indexer,col_indexer] = value instead

See the caveats in the documentation: https://pandas.pydata.org/pandas-docs/stable/user_guide/indexing.html#returning-a-view-versus-a-copy


predykcjareg_compare2.set_index(predykcjareg_compare2["Date_time"], inplace=True)
targeter2 = ['Predykcja', 'Share']
predykcjareg_series2 = darts.TimeSeries.from_dataframe(predykcjareg_compare2[targeter2])


fig, ax = plt.subplots(figsize=(24, 12))

predykcjareg_series2["Share"].plot(ax=ax, label='test')
predykcjareg_series2["Predykcja"].plot(ax=ax, label='forecast')

ax.set_xlabel('Date_time')
ax.set_ylabel('Share')
plt.show()


lgbm.plot_importance(modelreg2, max_num_features=40, figsize=(10,6))
plt.show()


Regressor2 = Regressor.drop(["Day_of_week","Date_time", "Share", "Audience"], axis=1)


threshold = 10
less_important_features = np.where(modelreg.feature_importances_ < threshold)[0]
high_important_features = np.where(modelreg.feature_importances_ > threshold)[0]


less_important_features

array([  0,   1,   2,   3,   4,   5,   6,   7,   8,   9,  10,  11,  12,
        13,  14,  15,  17,  18,  19,  20,  22,  23,  24,  25,  26,  27,
        28,  29,  30,  31,  32,  33,  34,  35,  36,  37,  38,  39,  40,
        41,  42,  43,  44,  46,  49,  50,  51,  52,  53,  54,  55,  56,
        57,  58,  59,  60,  61,  62,  65,  66,  67,  68,  70,  71,  72,
        73,  74,  75,  76,  77,  78,  80,  81,  82,  83,  84,  85,  86,
        87,  88,  89,  90,  91,  92,  93,  94,  95,  96,  97,  98,  99,
       100, 101, 102, 103, 104, 105, 106, 107, 108, 109, 110, 111, 112,
       113, 114, 115, 116, 117, 118, 119, 120, 121, 123, 124, 125, 126,
       127, 128, 129, 130, 131, 133, 134, 136, 137, 138, 139, 140, 141,
       142, 143, 144, 146, 147, 148, 149, 150, 151, 152, 153, 154, 155,
       156, 159, 160, 161, 162, 163, 165, 166, 167, 168, 169, 170, 171,
       172, 173, 174, 176, 177, 178, 179, 180, 181, 182, 183, 184, 185,
       187, 188, 189, 190, 191, 192, 193, 195, 196, 199, 200, 202, 203,
       204, 205, 206, 207, 208, 210, 211, 212, 213, 214, 215, 216, 218,
       219, 220, 222, 223, 224, 225, 226, 227, 228, 229, 230, 238],
      dtype=int64)


high_important_features

array([ 16,  45,  47,  48,  63,  64,  69,  79, 132, 135, 145, 157, 164,
       175, 186, 197, 198, 201, 209, 221, 231, 232, 233, 234, 235, 236,
       237, 239, 240, 241, 242, 243, 244, 245, 246, 247, 248, 249, 250,
       251, 252, 253], dtype=int64)


Regressor2.drop(Regressor2.columns[less_important_features], axis=1, inplace=True)


Regressor2['Audience'] = Regressor['Audience']
Regressor2['Share'] = Regressor['Share']
Regressor2['Date_time'] = Regressor['Date_time']


Regressor2.set_index(Regressor2.columns[-1], inplace=True)


Regressor2


testfin = Regressor2['2022-10-21':] # podział testowych danych na ostatnie dwa miesiące, gdyż taki okres będzie potrzebny do predykcji potem
trainfin = Regressor2[:'2022-10-20']
featuresdffin = Regressor2.drop(['Share', 'Audience'], axis = 1) #zrobić listę kolumn

featuresfin = list(featuresdffin.columns)
targetfin = ['Share', 'Audience']
target_ts_testfin = darts.TimeSeries.from_dataframe(testfin[targetfin])
feature_ts_testfin = darts.TimeSeries.from_dataframe(Regressor2[featuresfin])
target_ts_trainfin = darts.TimeSeries.from_dataframe(trainfin[targetfin])
feature_ts_trainfin = darts.TimeSeries.from_dataframe(trainfin[featuresfin])


model3 = LightGBMModel(
        lags=52,
        lags_past_covariates=8,
        lags_future_covariates=(16,8), 
        output_chunk_length=len(target_ts_testfin),
        add_encoders=None,
        likelihood=None,
        quantiles=None,
        random_state=None,
        )


model3.fit(target_ts_trainfin, feature_ts_trainfin, feature_ts_testfin)


#with open('model3_fut_cov.pkl', 'wb') as f3:
 #   pickle.dump(model3, f3)


with open('model3_fut_cov.pkl', 'rb') as f3:
    model3 = pickle.load(f3)


predykcja3 = model3.predict(len(target_ts_testfin), target_ts_trainfin, feature_ts_trainfin, feature_ts_testfin)


fig, ax = plt.subplots(figsize=(24, 12))

target_ts_testfin['Audience'].plot(ax=ax, label='test')
predykcja3['Audience'].plot(ax=ax, label='forecast')

ax.set_xlabel('Date_time')
ax.set_ylabel('Audience')
plt.show()


predykcja3df = predykcja3.pd_dataframe()
target_ts_testfindf = target_ts_testfin.pd_dataframe()


Compare3 = pd.merge(target_ts_testfindf['Audience'], predykcja3df['Audience'], left_index=True, right_on='Date_time', how='left')
Compare3.tail(60)


fig, ax = plt.subplots(figsize=(24, 12))

target_ts_test['Audience'][-124:].plot(ax=ax, label='test')
predykcja2['Audience'][-124:].plot(ax=ax, label='forecast fut cov')
predykcja['Audience'][-124:].plot(ax=ax, label='forecast pst cov')
predykcja3['Audience'][-124:].plot(ax=ax, label='forecast -feature')
predykcjareg_series["Predykcja"][-124:].plot(ax=ax, label='forecast regressor')

ax.set_xlabel('Date_time')
ax.set_ylabel('Audience')
plt.show()


fig, ax = plt.subplots(figsize=(24, 12))

target_ts_test['Audience'][248:-124].plot(ax=ax, label='test')
predykcja2['Audience'][248:-124].plot(ax=ax, label='forecast fut cov')
predykcja['Audience'][248:-124].plot(ax=ax, label='forecast pst cov')
predykcja3['Audience'][248:-124].plot(ax=ax, label='forecast -feature')
predykcjareg_series["Predykcja"][248:-124].plot(ax=ax, label='forecast regressor')

ax.set_xlabel('Date_time')
ax.set_ylabel('Audience')
plt.show()


fig, ax = plt.subplots(figsize=(24, 12))

target_ts_test['Audience'][124:248].plot(ax=ax, label='test')
predykcja2['Audience'][124:248].plot(ax=ax, label='forecast fut cov')
predykcja['Audience'][124:248].plot(ax=ax, label='forecast pst cov')
predykcja3['Audience'][124:248].plot(ax=ax, label='forecast -feature')
predykcjareg_series["Predykcja"][124:248].plot(ax=ax, label='forecast regressor')

ax.set_xlabel('Date_time')
ax.set_ylabel('Audience')
plt.show()


fig, ax = plt.subplots(figsize=(24, 12))

target_ts_test['Audience'][:124].plot(ax=ax, label='test')
predykcja2['Audience'][:124].plot(ax=ax, label='forecast fut cov')
predykcja['Audience'][:124].plot(ax=ax, label='forecast pst cov')
predykcja3['Audience'][:124].plot(ax=ax, label='forecast -feature')
predykcjareg_series["Predykcja"][:124].plot(ax=ax, label='forecast regressor')

ax.set_xlabel('Date_time')
ax.set_ylabel('Audience')
plt.show()


# porówananie metryk liczbowych w tabelce


data_val = pd.read_csv("Zbior walidacyjny.txt")


data_val = pd.DataFrame(data_val)
data_val


data_val["Date"] = pd.to_datetime(data_val["Date"])
data_val.insert(1, "Start_time_real", "") 
data_val.insert(1, "Date_real", "")


for i in data_val.index:
    if data_val["Start time"][i][:2] == '24':
        a = ("00" + data_val["Start time"][i][2:])
        b = data_val["Date"][i] + timedelta(days=1)
        data_val.at[i, 'Start_time_real'] = a
        data_val.at[i, 'Date_real'] = b
    elif data_val["Start time"][i][:2] == '25':
        a = ("01" + data_val["Start time"][i][2:])
        b = data_val["Date"][i] + timedelta(days=1)
        data_val.at[i, 'Start_time_real'] = a
        data_val.at[i, 'Date_real'] = b       
    else:
        a = data_val["Start time"][i]
        b = data_val["Date"][i]
        data_val.at[i, 'Start_time_real'] = a
        data_val.at[i, 'Date_real'] = b


data_val['Date_real'] = pd.to_datetime(data_val['Date_real'])
data_val["Date_time"] = pd.to_datetime(data_val['Date_real'] + pd.to_timedelta(data_val["Start_time_real"]))


work_val = data_val


work_val = work_val.drop('Date', axis = 1)
work_val = work_val.drop('Start time', axis = 1)
work_val = work_val.rename(columns={"Channel Group ...": "Channel", "Prog Share": "Share", "Programme Title": "Title", "'000 (avg)": "Audience"})


work_val["Title"].str.contains("Blad w emisji").value_counts()

False    29487
True         2
Name: Title, dtype: int64


work_val = work_val[work_val["Title"].str.contains("Blad w emisji") == False]  # czyszczenie danych z błędów emisyjnych


work_val["Title"].str.contains("Blad w emisji").value_counts()

False    29487
Name: Title, dtype: int64


History_val = work_val[work_val["Channel"] == "History"] 
History_val = History_val.drop(['Start_time_real','Channel'], axis = 1)
History_val


History_val_agg = History_val[['Date_time', 'Share', 'Title']]
df_val_agg = History_val_agg.groupby([pd.Grouper(key='Date_time', freq='3H'), 'Title'])['Share'].mean().reset_index()

time_index_val = pd.date_range(start=df_val_agg['Date_time'].min(), end=df_val_agg['Date_time'].max(), freq='3H')
time_df_val = pd.DataFrame(index=time_index_val)

df_val_agg.set_index('Date_time', inplace=True)

program_val_df = pd.get_dummies(df_val_agg['Title'])


channel_val_avg_shr = work_val.groupby('Channel')['Share'].mean().reset_index() 
channel_val_avg_shr = pd.DataFrame(channel_val_avg_shr)


work_val_agg = work_val[['Date_time', 'Share', 'Title', 'Channel']]
work_val_agg = work_val_agg.groupby([pd.Grouper(key='Date_time', freq='3H'), 'Title', 'Channel'])['Share'].mean().reset_index()
work_val_agg1 = work_val_agg.drop(['Share'], axis = 1)
work_val_agg_merged = pd.merge(work_val_agg1, heat_titles, on='Title')

work_val_agg_merged = work_val_agg_merged.drop(['Share', 'Title'], axis = 1) 
work_val_agg_grouped = work_val_agg_merged.groupby([pd.Grouper(key='Date_time', freq='3H'), pd.Grouper(key='Channel')])['Heat'].mean().reset_index()
pivot_val_chdf = work_val_agg_grouped.pivot(index='Date_time', columns='Channel', values='Heat')

for i in channel_val_avg_shr.index:
    if (channel_val_avg_shr['Channel'][i] == "History") or (channel_val_avg_shr['Share'][i] < 0.1):
        name_val = channel_val_avg_shr['Channel'][i]
        pivot_val_chdf = pivot_val_chdf.drop(str(name_val), axis = 1)  
        
pivot_val_chdf


History_val_granulated = History_val[['Date_time', 'Share', 'Audience']]
History_val_granulated.set_index('Date_time', inplace=True)
History_val_granulated = History_val_granulated.resample('3H').mean()
History_val_granulated['Share'].fillna(0, inplace=True)
History_val_granulated['Audience'].fillna(0, inplace=True)


result_val = pd.merge(time_df_val, program_val_df, left_index=True, right_on='Date_time', how='left') 
result_val = result_val.fillna(0)
result_val = result_val.groupby(level=0).sum()
result_val


ShareOHE_val_temp = pd.merge(History_val_granulated, result_val, left_index=True, right_on='Date_time', how='left') 
ShareOHE_val = pd.merge(ShareOHE_val_temp, pivot_val_chdf, left_index = True, right_on='Date_time')
ShareOHE_val = ShareOHE_val.fillna(0)
ShareOHE_val = ShareOHE_val[1:] # usunięcie wiersza, który będzie się powtarzał przy łączeniu ramek
ShareOHE_val = ShareOHE_val[:-55] # usunięcie ostatnich wierszy z uwagi na zbyt daleką predykcja (model2 ma horyzont 497)


ShareOHE['Audience'].plot(label='train')
ShareOHE_val['Audience'].plot(label='test')

<AxesSubplot:xlabel='Date_time'>


featuresdf_val = ShareOHE.drop(['Share', 'Audience'], axis = 1) #zrobić listę kolumn

features_val = list(featuresdf_val.columns)


ShareOHE_combined = pd.concat([ShareOHE, ShareOHE_val])
ShareOHE_combined = ShareOHE_combined.fillna(0)
ShareOHE_combined


target_ts_val = darts.TimeSeries.from_dataframe(ShareOHE_val[target])
feature_ts_val = darts.TimeSeries.from_dataframe(ShareOHE_combined[features])
target_ts_past = darts.TimeSeries.from_dataframe(ShareOHE[target])
feature_ts_past = darts.TimeSeries.from_dataframe(ShareOHE[features])


predykcja_val = model2.predict(len(target_ts_val), target_ts_past, feature_ts_past, feature_ts_val)


fig, ax = plt.subplots(figsize=(24, 12))

target_ts_val['Audience'].plot(ax=ax, label='validation')
predykcja_val['Audience'].plot(ax=ax, label='forecast')

ax.set_xlabel('Date_time')
ax.set_ylabel('Audience')
plt.show()


fig, ax = plt.subplots(figsize=(24, 12))

target_ts_val['Audience'][-124:].plot(ax=ax, label='validation')
predykcja_val['Audience'][-124:].plot(ax=ax, label='forecast')

ax.set_xlabel('Date_time')
ax.set_ylabel('Audience')
plt.show()


fig, ax = plt.subplots(figsize=(24, 12))

target_ts_val['Audience'][248:-124].plot(ax=ax, label='validation')
predykcja_val['Audience'][248:-124].plot(ax=ax, label='forecast')

ax.set_xlabel('Date_time')
ax.set_ylabel('Audience')
plt.show()


fig, ax = plt.subplots(figsize=(24, 12))

target_ts_val['Audience'][124:248].plot(ax=ax, label='validation')
predykcja_val['Audience'][124:248].plot(ax=ax, label='forecast')

ax.set_xlabel('Date_time')
ax.set_ylabel('Audience')
plt.show()


fig, ax = plt.subplots(figsize=(24, 12))

target_ts_val['Audience'][:124].plot(ax=ax, label='validation')
predykcja_val['Audience'][:124].plot(ax=ax, label='forecast')

ax.set_xlabel('Date_time')
ax.set_ylabel('Audience')
plt.show()


darts.metrics.metrics.mae(target_ts_val["Audience"], predykcja_val["Audience"], intersect=True)

4.564902648903264


darts.metrics.metrics.mse(target_ts_val["Audience"], predykcja_val["Audience"], intersect=True)

37.585639461775195


darts.metrics.r2_score(target_ts_val["Audience"], predykcja_val["Audience"])

0.31439550359429724


darts.metrics.metrics.rmse(target_ts_val["Audience"], predykcja_val["Audience"], intersect=True)

6.1307128020952995


predykcja_val_df = predykcja_val.pd_dataframe()
target_ts_val_df = target_ts_val.pd_dataframe()


Compare_val = pd.merge(target_ts_val_df['Audience'], predykcja_val_df['Audience'], left_index=True, right_on='Date_time', how='left')
Compare_val

	Channel Group ...	Date	Start time	Programme Title	'000 (avg)	Prog Share
0	History	2017-12-22	02:00:00	Wykute w ogniu	0.000	0.000
1	History	2017-12-22	02:27:00	Ludzie gor	0.000	0.000
2	History	2017-12-22	03:11:00	Gwiazdy lombardu	0.000	0.000
3	History	2017-12-22	03:32:00	Wojny magazynowe	3.234	0.499
4	History	2017-12-22	03:55:00	Wykute w ogniu	0.000	0.000
...	...	...	...	...	...	...
1403571	Discovery Historia	2022-12-21	22:00:00	Nowe zycie starych rzeczy	2.308	0.042
1403572	Discovery Historia	2022-12-21	22:30:00	Nowe zycie starych rzeczy	8.183	0.176
1403573	Discovery Historia	2022-12-21	23:00:00	Historia oreza	4.417	0.125
1403574	Discovery Historia	2022-12-21	24:00:00	Miecze z szalonej kuzni	2.930	0.141
1403575	Discovery Historia	2022-12-21	25:00:00	Miecze z szalonej kuzni	0.000	0.000

	Channel	Date_real	Start_time_real	Title	Audience	Share	Date_time	time_band	Day_of_week
0	History	2017-12-22	02:00:00	Wykute w ogniu	0.000	0.000	2017-12-22 02:00:00	band 0-3	Friday
1	History	2017-12-22	02:27:00	Ludzie gor	0.000	0.000	2017-12-22 02:27:00	band 0-3	Friday
2	History	2017-12-22	03:11:00	Gwiazdy lombardu	0.000	0.000	2017-12-22 03:11:00	band 3-6	Friday
3	History	2017-12-22	03:32:00	Wojny magazynowe	3.234	0.499	2017-12-22 03:32:00	band 3-6	Friday
4	History	2017-12-22	03:55:00	Wykute w ogniu	0.000	0.000	2017-12-22 03:55:00	band 3-6	Friday
...	...	...	...	...	...	...	...	...	...
1403571	Discovery Historia	2022-12-21	22:00:00	Nowe zycie starych rzeczy	2.308	0.042	2022-12-21 22:00:00	band 21-24	Wednesday
1403572	Discovery Historia	2022-12-21	22:30:00	Nowe zycie starych rzeczy	8.183	0.176	2022-12-21 22:30:00	band 21-24	Wednesday
1403573	Discovery Historia	2022-12-21	23:00:00	Historia oreza	4.417	0.125	2022-12-21 23:00:00	band 21-24	Wednesday
1403574	Discovery Historia	2022-12-22	00:00:00	Miecze z szalonej kuzni	2.930	0.141	2022-12-22 00:00:00	band 0-3	Thursday
1403575	Discovery Historia	2022-12-22	01:00:00	Miecze z szalonej kuzni	0.000	0.000	2022-12-22 01:00:00	band 0-3	Thursday

	Day_of_week	Audience
0	Friday	12.704358
1	Monday	12.799383
2	Saturday	14.789853
3	Sunday	21.480364
4	Thursday	12.963582
5	Tuesday	12.509920
6	Wednesday	12.604528

	Title	Share
7593	Uwaga mister baker	0.000000
662	Bo jestem geniuszem lorenza mazzetti	0.000000
4802	Odkryc prawde	0.000000
180	Afryka mojego dziecinstwa	0.000000
2731	Jestesmy w raju	0.000000
...	...	...
6009	Rozmiar xxl na sprzedaz	4.159000
611	Bilet na ksiezyc	4.616500
6353	Slady zbrodni	4.713333
599	Bierz forse i w nogi	4.762000
4068	Morderstwo w domku dla lalek	4.817000

	Title	Share	Heat
8663	Zwyciezcy i przegrani	0.021500	0
8664	Zwyciezcy nie umieraja opowiesc o ksiedz	0.000000	0
8665	Zwyczajny kraj	0.172667	0
8666	Zwykle rzeczy niezwykle wynalazki	0.281967	0
8667	Zwykle rzeczy niezwykle wynalazki sport	0.023000	0
8668	Zwykle zaburzenie	0.096800	0
8669	Zyc nie umierac	0.152889	0
8670	Zycia nie mozna zmarnowac	0.072000	0
8671	Zycie baltyku	0.668538	1
8672	Zycie bez happy endu	0.050375	0
8673	Zycie bez rak historia jessiki cox	0.127667	0
8674	Zycie lasu bogactwo runa i podszytu	0.278375	0
8675	Zycie lasu grzyby	0.128000	0
8676	Zycie lasu ptaki	0.317333	0
8677	Zycie lasu storczyki	0.376800	1
8678	Zycie na alasce	0.204025	0
8679	Zycie na cienkiej linie	0.210000	0
8680	Zycie na krawedzi	0.360552	1
8681	Zycie na tafli lodu	0.619409	1
8682	Zycie na trawlerach	1.018167	1
8683	Zycie nelsona mandeli	0.051563	0
8684	Zycie pierwotnych	0.310800	0
8685	Zycie po cudzie	0.094333	0
8686	Zycie po zagladzie ludzi	0.154544	0
8687	Zycie pod woda	0.000000	0
8688	Zycie przed oczami	1.480067	1
8689	Zycie przed toba	0.055000	0
8690	Zycie to cos pieknego	0.010000	0
8691	Zycie w blekicie	0.131200	0
8692	Zycie w drodze medycyna mongolskich step	0.170750	0
8693	Zycie w drodze ostatnie mobilne kino	0.113250	0
8694	Zycie w dziczy	0.480651	1
8695	Zycie w kosmosie	0.373857	1
8696	Zycie w piecdziesieciu stopniach	0.179200	0
8697	Zycie w powietrzu	0.146000	0
8698	Zycie w wiezieniu	0.560000	1
8699	Zycie z efektem wow!	0.494050	1
8700	Zycie za kolkiem	0.857437	1
8701	Zycie zaczyna sie od tanca	0.036000	0
8702	Zyciodajne rzeki polwyspu iberyjskiego	0.187029	0
8703	Zyciorys	0.034000	0
8704	Zyciorys jak z orwella	0.021000	0
8705	Zyczenia swiateczne	0.012140	0
8706	Zyczenie smierci v	0.742500	1
8707	Zydzi w powstaniu warszawskim	0.071500	0
8708	Zygmunt hubner gra z rzeczywistoscia	0.080167	0
8709	Zygmunt konieczny	0.000000	0
8710	Zyjac z tourettem	0.406857	1
8711	Zylem jak chcialem... gen. stanislaw ska	0.044511	0
8712	Zylismy tylko muzyka	0.018600	0
8713	Zyrafa swiat z wysoka	0.134545	0
8714	Zywe pomniki najstarsze drzewa w polsce	0.172000	0
8715	Zywiol	0.454833	1
8716	Zywioly atakuja	0.341500	1
8717	Zywioly grochowiaka	0.000000	0
8718	Zywioly ktore niosa smierc	0.118828	0
8719	Zywnosc 3.0	0.047500	0
8720	Zywnosc jutra	0.139694	0
8721	Zywot michala	0.203429	0
8722	Zza plotu	0.011889	0

Prognozowanie oglądalności dla stacji telewizyjnej z wykorzystaniem szeregów czasowych.¶

1. Wstęp¶

2. Wykorzystywane biblioteki i metody¶

3. Wczytywanie danych wraz z ich oczyszczaniem i wstępnym przygotowaniem do analizy¶

4. Grupowanie danych celem ustalenia potencjalnego wpływu na wyniki oglądalności¶

4.1 Sprawdzenie zrożnicowania średniej oglądalności wszystkich stacji w poszczególnych dniach.¶

4.2 Sprawdzenie zróżnicowania średniej oglądalności w zależności od pasma programowego (podział co 3 godziny).¶

4.3 Tworzenie ramki danych określającej dany tytuł jako heat (ponad przeciętna oglądalność).¶

4.4 Wydzielenie osobnej ramki dla kanału dedykowanego do predykcji.¶

5. Tworzenie ramki danych do modelowania szeregów czasowych ze wspołzmiennymi.¶

5.1 Stworzenie ramki danych z pozycjami heat'owymi dla kanałów o oglądalności powyżej 0.1 w trzygodzinnych pasmach programowych.¶

5.2 Łączenie średniej oglądalności i widowni programu History z tabelami przedstawiającymi, które tytuły były grane w danym paśmie i ze średnim pokryciem pozycjami heat'owymi w stacjach konkurencyjnych.¶

6. Modelowanie szeregu czasowego¶

6.2 Model darts LightGBM z wykorzystaniem tylko przeszłych współzmiennych (past covariates).¶

6.2.1 Metryki¶

6.3 Model darts LightGBM z wykorzystaniem przyszłych współzmiennych (future covariates).¶

6.3.1 Metryki¶

6.4 Model LGBMRegressor dla porównania i wyboru optymalnych feature'ów.¶

6.4.1 Metryki¶

6.5 Model darts LightGBM po odsianiu teorertycznie mniej istotnych feature'ów.¶

8. Podsumowanie i interpretacja wyników.¶

9. Weryfikacja najlepszego otrzymanego modelu na nowych danych telemetrycznych.¶

9.1 Wczytywanie ramówki na kolejne dwa miesiące i przygotowanie zbioru do weryfikacji modelu.¶

9.2 Walidacja modelu.¶

9.3 Podsumowanie walidacji modelu.¶

	Title	Share
0	#chicagogirl - facebookowa rewolucja	0.044000
1	#jestem m misfit	0.229278
2	#liniaprosta	0.294787
3	#piatkadlarysia	0.223189
4	#tvn pomagajmy sobie	0.299625
...	...	...
8718	Zywioly ktore niosa smierc	0.118828
8719	Zywnosc 3.0	0.047500
8720	Zywnosc jutra	0.139694
8721	Zywot michala	0.203429
8722	Zza plotu	0.011889

	Channel	Share
0	BBC Brit	0.094470
1	BBC Earth	0.159274
2	Discovery Historia	0.112552
3	Discovery Science	0.275614
4	Discovery pl	0.629033
5	Fokus TV	1.159488
6	History	0.429808
7	History2 [H2]	0.132355
8	National Geographic	0.420973
9	Polsat Games	0.040729
10	Polsat Play	0.602589
11	Polsat Viasat Explore	0.187947
12	Polsat Viasat History	0.257810
13	TVN Turbo	0.765775
14	TVP Dokument	0.091259
15	WP	0.681939
16	Zoom TV	0.555794

	count	mean	std	min	25%	50%	75%	max
Channel
BBC Brit	46815.0	0.094470	0.240331	0.0	0.000	0.010	0.091	6.565
BBC Earth	48491.0	0.159274	0.322483	0.0	0.000	0.042	0.180	7.475
Discovery Historia	59143.0	0.112552	0.273309	0.0	0.000	0.014	0.108	6.155
Discovery Science	71866.0	0.275614	0.452456	0.0	0.016	0.128	0.342	8.597
Discovery pl	51919.0	0.629033	0.686551	0.0	0.165	0.444	0.860	8.484
Fokus TV	122754.0	1.159488	1.193825	0.0	0.454	0.892	1.483	18.031
History	74022.0	0.429808	0.643947	0.0	0.034	0.234	0.569	12.340
History2 [H2]	47402.0	0.132355	0.316290	0.0	0.000	0.023	0.144	7.229
National Geographic	50130.0	0.420973	0.542777	0.0	0.099	0.269	0.541	13.823
Polsat Games	100923.0	0.040729	0.161432	0.0	0.000	0.000	0.000	8.609
Polsat Play	157380.0	0.602589	0.698619	0.0	0.107	0.421	0.856	12.673
Polsat Viasat Explore	60236.0	0.187947	0.423695	0.0	0.000	0.042	0.196	9.659
Polsat Viasat History	62573.0	0.257810	0.464621	0.0	0.000	0.100	0.308	11.020
TVN Turbo	151995.0	0.765775	0.762277	0.0	0.184	0.589	1.125	14.860
TVP Dokument	24629.0	0.091259	0.168507	0.0	0.000	0.023	0.112	3.662
WP	127704.0	0.681939	0.921885	0.0	0.121	0.409	0.852	14.833
Zoom TV	143756.0	0.555794	0.850709	0.0	0.019	0.290	0.715	19.950

	Date_time	Title	Share
0	2017-12-22 00:00:00	Ludzie gor	0.000
1	2017-12-22 00:00:00	Wykute w ogniu	0.000
2	2017-12-22 03:00:00	Auto-reanimacja detroit	0.000
3	2017-12-22 03:00:00	Gwiazdy lombardu	0.000
4	2017-12-22 03:00:00	Poznaj history	0.000
...	...	...	...
39987	2022-12-21 21:00:00	Starozytni kosmici	0.455
39988	2022-12-21 21:00:00	Tajemnice skryte w dzungli	0.306
39989	2022-12-22 00:00:00	Jak to wyjasnic	0.212
39990	2022-12-22 00:00:00	Ludzie gor	1.248
39991	2022-12-22 00:00:00	Starozytni kosmici	0.081

Channel	BBC Earth	Discovery Historia	Discovery Science	Discovery pl	Fokus TV	History2 [H2]	National Geographic	Polsat Play	Polsat Viasat Explore	Polsat Viasat History	TVN Turbo	WP	Zoom TV
Date_time
2017-12-22 00:00:00	0.0	0.250000	0.250000	0.666667	1.00	0.500000	1.000000	1.000000	0.0	0.000000	1.0	1.00	0.666667
2017-12-22 03:00:00	0.0	0.000000	0.750000	0.833333	1.00	0.000000	1.000000	0.800000	0.0	0.571429	0.8	1.00	0.400000
2017-12-22 06:00:00	0.0	0.000000	0.428571	0.750000	1.00	0.333333	1.000000	0.833333	0.0	0.000000	0.8	1.00	0.000000
2017-12-22 09:00:00	0.0	0.250000	0.600000	1.000000	0.75	0.000000	0.666667	1.000000	0.0	0.000000	1.0	0.75	0.333333
2017-12-22 12:00:00	0.0	0.333333	0.500000	1.000000	1.00	0.333333	1.000000	0.750000	0.0	0.000000	1.0	1.00	0.600000
...	...	...	...	...	...	...	...	...	...	...	...	...	...
2022-12-21 12:00:00	0.0	0.000000	0.500000	1.000000	1.00	0.250000	1.000000	1.000000	0.0	0.500000	1.0	1.00	1.000000
2022-12-21 15:00:00	0.0	0.500000	1.000000	1.000000	1.00	0.333333	1.000000	1.000000	0.0	0.500000	1.0	0.75	1.000000
2022-12-21 18:00:00	0.0	0.000000	0.000000	1.000000	1.00	0.000000	0.750000	1.000000	0.0	0.000000	1.0	1.00	1.000000
2022-12-21 21:00:00	0.0	0.000000	0.000000	1.000000	1.00	0.000000	0.500000	1.000000	0.0	0.000000	1.0	1.00	1.000000
2022-12-22 00:00:00	0.0	0.000000	0.500000	1.000000	1.00	0.500000	1.000000	1.000000	0.0	0.000000	1.0	1.00	1.000000

	Share	Audience	#liniaprosta	Ale bryka	Alone: przetrwac najdluzej	Amerykanscy naprawiacze	Amerykanscy zbieracze	Amerykanska ksiega tajemnic	Amerykanski rozpruwacz w londynie	Atlantyda odnaleziona	...	Discovery pl	Fokus TV	History2 [H2]	National Geographic	Polsat Play	Polsat Viasat Explore	Polsat Viasat History	TVN Turbo	WP	Zoom TV
Date_time
2022-12-14 15:00:00	0.543000	21.272250	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	...	1.000000	1.000000	0.333333	1.000000	1.0	0.000000	0.000000	1.000000	0.833333	1.000000
2022-12-14 18:00:00	0.629667	37.457000	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	...	1.000000	1.000000	0.000000	0.666667	1.0	0.000000	0.000000	1.000000	1.000000	1.000000
2022-12-14 21:00:00	0.205750	6.754500	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	...	1.000000	1.000000	0.000000	0.600000	1.0	0.000000	0.000000	1.000000	1.000000	1.000000
2022-12-15 00:00:00	0.144400	2.417400	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	...	1.000000	1.000000	0.333333	1.000000	1.0	0.000000	0.000000	1.000000	1.000000	1.000000
2022-12-15 03:00:00	0.010250	0.047250	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	...	1.000000	1.000000	0.000000	0.500000	1.0	0.000000	0.333333	0.833333	1.000000	1.000000
2022-12-15 06:00:00	0.131667	2.519167	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	...	0.666667	1.000000	0.333333	0.750000	1.0	0.000000	0.250000	1.000000	0.750000	0.333333
2022-12-15 09:00:00	0.241667	5.559667	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	...	1.000000	1.000000	0.000000	0.750000	1.0	0.000000	0.250000	1.000000	1.000000	0.750000
2022-12-15 12:00:00	0.548714	13.815857	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	...	1.000000	1.000000	0.333333	1.000000	1.0	0.000000	0.000000	1.000000	1.000000	1.000000
2022-12-15 15:00:00	0.291250	11.399250	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	...	1.000000	1.000000	0.333333	1.000000	1.0	0.000000	0.000000	1.000000	0.750000	1.000000
2022-12-15 18:00:00	0.191333	11.551333	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	...	1.000000	1.000000	0.000000	0.750000	1.0	0.000000	0.333333	1.000000	0.800000	1.000000
2022-12-15 21:00:00	0.153000	5.465000	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	...	0.750000	1.000000	0.333333	1.000000	1.0	0.000000	0.000000	1.000000	1.000000	1.000000
2022-12-16 00:00:00	0.125600	1.119200	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	...	1.000000	1.000000	0.000000	1.000000	1.0	0.000000	0.250000	1.000000	1.000000	1.000000
2022-12-16 03:00:00	0.169000	0.997250	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	...	1.000000	1.000000	0.000000	0.666667	1.0	0.000000	0.500000	0.750000	1.000000	1.000000
2022-12-16 06:00:00	0.197000	2.477833	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	...	1.000000	1.000000	0.333333	0.750000	1.0	0.000000	0.250000	1.000000	0.750000	0.166667
2022-12-16 09:00:00	0.508667	11.551167	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	...	0.750000	1.000000	0.000000	0.750000	1.0	0.000000	0.000000	1.000000	1.000000	0.750000
2022-12-16 12:00:00	0.278000	7.380714	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	...	1.000000	1.000000	0.333333	1.000000	1.0	0.000000	0.333333	1.000000	1.000000	1.000000
2022-12-16 15:00:00	0.384750	14.283250	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	...	1.000000	1.000000	0.333333	1.000000	1.0	0.000000	0.000000	1.000000	1.000000	1.000000
2022-12-16 18:00:00	0.331333	19.037667	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	...	1.000000	1.000000	0.000000	0.750000	1.0	1.000000	0.000000	1.000000	1.000000	1.000000
2022-12-16 21:00:00	0.179250	7.866000	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	...	0.750000	1.000000	0.333333	0.666667	1.0	0.000000	0.333333	1.000000	1.000000	0.666667
2022-12-17 00:00:00	0.281000	4.419000	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	...	1.000000	1.000000	0.333333	1.000000	1.0	0.000000	0.250000	1.000000	1.000000	1.000000
2022-12-17 03:00:00	0.984167	4.528833	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	...	1.000000	1.000000	0.333333	0.666667	1.0	0.000000	0.333333	0.833333	1.000000	1.000000
2022-12-17 06:00:00	0.907800	7.926800	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	...	1.000000	0.750000	0.250000	1.000000	1.0	0.000000	0.250000	1.000000	1.000000	0.400000
2022-12-17 09:00:00	0.119286	3.884714	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	...	1.000000	0.800000	0.333333	1.000000	1.0	0.000000	0.000000	1.000000	1.000000	0.666667
2022-12-17 12:00:00	0.301625	11.627000	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	...	1.000000	1.000000	0.333333	1.000000	1.0	0.000000	0.000000	1.000000	1.000000	1.000000
2022-12-17 15:00:00	0.122333	6.837333	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	...	1.000000	1.000000	0.000000	0.666667	1.0	0.000000	0.000000	1.000000	1.000000	0.750000
2022-12-17 18:00:00	0.114333	7.166333	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	...	0.666667	1.000000	0.000000	0.000000	1.0	0.000000	0.000000	1.000000	0.666667	1.000000
2022-12-17 21:00:00	0.109500	4.158500	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	...	0.750000	1.000000	0.500000	0.000000	1.0	0.000000	0.000000	1.000000	1.000000	1.000000
2022-12-18 00:00:00	1.099000	13.705500	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	...	0.000000	1.000000	0.333333	1.000000	1.0	0.000000	0.000000	1.000000	1.000000	1.000000
2022-12-18 03:00:00	0.379600	1.599600	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	...	1.000000	1.000000	0.500000	1.000000	1.0	0.000000	0.600000	0.857143	1.000000	1.000000
2022-12-18 06:00:00	0.084200	1.475000	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	...	0.333333	0.750000	0.000000	1.000000	1.0	0.000000	0.333333	1.000000	0.750000	0.400000
2022-12-18 09:00:00	0.227667	9.744667	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	...	1.000000	0.666667	0.000000	0.800000	1.0	0.000000	0.000000	1.000000	1.000000	0.666667
2022-12-18 12:00:00	0.164000	7.983500	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	...	0.750000	1.000000	0.333333	1.000000	1.0	0.000000	0.000000	1.000000	1.000000	1.000000
2022-12-18 15:00:00	0.187000	13.259333	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	...	1.000000	1.000000	0.000000	0.666667	1.0	0.000000	0.000000	1.000000	0.750000	1.000000
2022-12-18 18:00:00	0.157250	12.743500	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	...	1.000000	1.000000	0.000000	0.666667	1.0	0.000000	0.000000	1.000000	1.000000	1.000000
2022-12-18 21:00:00	0.198750	8.042000	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	...	0.666667	1.000000	0.000000	1.000000	1.0	0.000000	0.000000	1.000000	1.000000	1.000000
2022-12-19 00:00:00	0.205750	1.867500	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	...	1.000000	1.000000	0.000000	0.666667	1.0	0.000000	0.000000	1.000000	1.000000	1.000000
2022-12-19 03:00:00	0.555333	2.890000	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	...	0.500000	1.000000	0.000000	1.000000	1.0	0.000000	0.666667	0.800000	1.000000	1.000000
2022-12-19 06:00:00	0.195333	3.124333	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	...	1.000000	1.000000	0.000000	0.666667	1.0	0.000000	0.666667	1.000000	0.800000	0.333333
2022-12-19 09:00:00	0.610500	13.616000	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	...	1.000000	1.000000	0.000000	0.666667	1.0	0.000000	0.000000	1.000000	1.000000	0.666667
2022-12-19 12:00:00	0.970571	24.233571	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	...	1.000000	1.000000	0.000000	1.000000	1.0	0.000000	0.333333	1.000000	1.000000	1.000000
2022-12-19 15:00:00	0.665750	25.171000	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	...	1.000000	1.000000	0.333333	1.000000	1.0	0.000000	0.500000	1.000000	0.750000	1.000000
2022-12-19 18:00:00	0.138000	8.418333	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	...	1.000000	1.000000	0.000000	0.750000	1.0	0.000000	0.000000	1.000000	1.000000	1.000000
2022-12-19 21:00:00	0.303000	14.754000	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	...	0.500000	1.000000	0.500000	1.000000	1.0	0.333333	0.666667	1.000000	1.000000	1.000000
2022-12-20 00:00:00	0.152000	1.484400	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	...	1.000000	1.000000	0.333333	1.000000	1.0	0.000000	0.000000	1.000000	1.000000	1.000000
2022-12-20 03:00:00	0.038000	0.231250	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	...	1.000000	1.000000	0.333333	0.500000	1.0	0.000000	0.500000	0.750000	1.000000	1.000000
2022-12-20 06:00:00	0.043000	0.822333	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	...	1.000000	1.000000	0.333333	0.666667	1.0	0.000000	0.500000	1.000000	0.750000	0.333333
2022-12-20 09:00:00	0.525333	11.786333	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	...	1.000000	1.000000	0.333333	0.750000	1.0	0.000000	0.000000	1.000000	1.000000	0.750000
2022-12-20 12:00:00	0.751571	18.820857	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	...	1.000000	1.000000	0.500000	1.000000	1.0	0.000000	0.333333	1.000000	1.000000	1.000000
2022-12-20 15:00:00	0.230000	7.818250	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	...	1.000000	1.000000	0.333333	1.000000	1.0	0.000000	0.333333	1.000000	0.750000	1.000000
2022-12-20 18:00:00	0.174667	10.428333	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	...	1.000000	1.000000	0.000000	0.750000	1.0	0.000000	0.000000	1.000000	1.000000	1.000000
2022-12-20 21:00:00	0.147750	6.036250	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	...	0.666667	1.000000	0.333333	1.000000	1.0	0.000000	0.666667	1.000000	1.000000	1.000000
2022-12-21 00:00:00	1.124800	9.992600	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	...	1.000000	1.000000	0.333333	1.000000	1.0	0.000000	0.500000	1.000000	1.000000	1.000000
2022-12-21 03:00:00	0.272750	1.094000	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	...	1.000000	1.000000	0.333333	0.500000	1.0	0.000000	0.500000	0.833333	1.000000	1.000000
2022-12-21 06:00:00	0.650500	10.044667	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	...	1.000000	1.000000	0.250000	0.750000	1.0	0.000000	0.333333	1.000000	0.750000	0.333333
2022-12-21 09:00:00	0.143833	3.211167	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	...	1.000000	1.000000	0.333333	0.750000	1.0	0.000000	0.000000	1.000000	1.000000	0.750000
2022-12-21 12:00:00	0.523857	13.609429	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	...	1.000000	1.000000	0.250000	1.000000	1.0	0.000000	0.500000	1.000000	1.000000	1.000000
2022-12-21 15:00:00	0.744250	26.405750	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	...	1.000000	1.000000	0.333333	1.000000	1.0	0.000000	0.500000	1.000000	0.750000	1.000000
2022-12-21 18:00:00	0.199333	11.880333	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	...	1.000000	1.000000	0.000000	0.750000	1.0	0.000000	0.000000	1.000000	1.000000	1.000000
2022-12-21 21:00:00	0.366250	16.568750	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	...	1.000000	1.000000	0.000000	0.500000	1.0	0.000000	0.000000	1.000000	1.000000	1.000000
2022-12-22 00:00:00	0.513667	5.893667	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	...	1.000000	1.000000	0.500000	1.000000	1.0	0.000000	0.000000	1.000000	1.000000	1.000000

	Audience_x	Audience_y
Date_time
2022-10-21 00:00:00	22.497600	6.578962
2022-10-21 03:00:00	4.165500	2.411158
2022-10-21 06:00:00	2.827667	4.951435
2022-10-21 09:00:00	23.670167	11.122972
2022-10-21 12:00:00	42.930714	17.846102
2022-10-21 15:00:00	26.768250	14.495430
2022-10-21 18:00:00	21.017667	11.853923
2022-10-21 21:00:00	8.236250	10.154472
2022-10-22 00:00:00	6.115200	3.495582
2022-10-22 03:00:00	3.325750	2.153462
2022-10-22 06:00:00	9.128600	5.299115
2022-10-22 09:00:00	15.286667	13.880580
2022-10-22 12:00:00	11.177286	18.113552
2022-10-22 15:00:00	16.250750	14.060911
2022-10-22 18:00:00	8.910667	11.154516
2022-10-22 21:00:00	4.567667	13.579316
2022-10-23 00:00:00	1.450000	4.796421
2022-10-23 03:00:00	1.424800	3.210697
2022-10-23 06:00:00	4.655000	6.527127
2022-10-23 09:00:00	16.935000	20.417028
2022-10-23 12:00:00	21.104600	18.644642
2022-10-23 15:00:00	5.667333	12.377981
2022-10-23 18:00:00	22.253000	10.390696
2022-10-23 21:00:00	9.043250	11.358614
2022-10-24 00:00:00	2.207750	3.789730
2022-10-24 03:00:00	4.296333	3.930510
2022-10-24 06:00:00	4.764000	4.164865
2022-10-24 09:00:00	14.300167	14.191240
2022-10-24 12:00:00	29.858857	15.606195
2022-10-24 15:00:00	16.200250	13.193949
2022-10-24 18:00:00	3.084667	13.529369
2022-10-24 21:00:00	16.453500	10.972504
2022-10-25 00:00:00	15.541000	3.812802
2022-10-25 03:00:00	3.000750	2.130084
2022-10-25 06:00:00	2.043167	5.398361
2022-10-25 09:00:00	5.150167	8.785095
2022-10-25 12:00:00	30.346429	15.894498
2022-10-25 15:00:00	22.833500	12.389926
2022-10-25 18:00:00	19.528000	11.597486
2022-10-25 21:00:00	27.955500	11.094284
2022-10-26 00:00:00	0.292000	3.951129
2022-10-26 03:00:00	0.302000	2.196523
2022-10-26 06:00:00	4.237833	5.860232
2022-10-26 09:00:00	25.682000	10.567462
2022-10-26 12:00:00	31.741429	14.372916
2022-10-26 15:00:00	14.825250	12.679296
2022-10-26 18:00:00	8.404667	9.039912
2022-10-26 21:00:00	18.631250	15.744884
2022-10-27 00:00:00	1.312600	4.799364
2022-10-27 03:00:00	0.399500	1.983686
2022-10-27 06:00:00	1.537500	4.351796
2022-10-27 09:00:00	3.355000	8.858796
2022-10-27 12:00:00	11.596143	14.640186
2022-10-27 15:00:00	18.575750	15.246026
2022-10-27 18:00:00	6.374000	9.348100
2022-10-27 21:00:00	6.257000	9.189762
2022-10-28 00:00:00	1.627600	4.608043
2022-10-28 03:00:00	0.228500	2.185046
2022-10-28 06:00:00	11.155500	4.778248
2022-10-28 09:00:00	13.957667	10.529174

	Date_time	Audience	Predykcja
14112	2022-10-21 00:00:00	22.497600	4.754351
14113	2022-10-21 03:00:00	4.165500	2.726264
14114	2022-10-21 06:00:00	2.827667	4.693885
14115	2022-10-21 09:00:00	23.670167	11.373406
14116	2022-10-21 12:00:00	42.930714	15.167319
...	...	...	...
14604	2022-12-21 12:00:00	13.609429	21.072565
14605	2022-12-21 15:00:00	26.405750	19.297015
14606	2022-12-21 18:00:00	11.880333	18.999243
14607	2022-12-21 21:00:00	16.568750	14.521220
14608	2022-12-22 00:00:00	5.893667	4.619446

	Date_time	Share	Predykcja
time
14112	2022-10-21 00:00:00	2.280000	0.484746
14113	2022-10-21 03:00:00	1.141500	0.530190
14114	2022-10-21 06:00:00	0.166333	0.375726
14115	2022-10-21 09:00:00	1.153667	0.498783
14116	2022-10-21 12:00:00	2.023429	0.584274
...	...	...	...
14604	2022-12-21 12:00:00	0.523857	0.840509
14605	2022-12-21 15:00:00	0.744250	0.529757
14606	2022-12-21 18:00:00	0.199333	0.267945
14607	2022-12-21 21:00:00	0.366250	0.274163
14608	2022-12-22 00:00:00	0.513667	0.627100

	Bitwy_magazynowe	Brytyjskie_gwiazdy_lombardu	Geniusze_futbolu	Gwiazdy_lombardu	Gwiazdy_lombardu_rpa	Klatwa_wyspy_debow	Kod_ukryty_w_biblii	Korea_polnocna_tajemnice_rezimu	Ludzie_gor	Ostatni_papiez_	...	Saturday	Sunday	Thursday	Tuesday	Wednesday	Year	Month	Week	Audience	Share
Date_time
2017-12-22 00:00:00	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	1.0	0.0	...	0	0	0	0	0	2017	12	51	0.000000	0.000000
2017-12-22 03:00:00	0.0	0.0	0.0	1.0	0.0	0.0	0.0	0.0	0.0	0.0	...	0	0	0	0	0	2017	12	51	0.539000	0.083167
2017-12-22 06:00:00	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	1.0	0.0	...	0	0	0	0	0	2017	12	51	0.859000	0.083750
2017-12-22 09:00:00	0.0	0.0	0.0	1.0	0.0	0.0	0.0	0.0	1.0	0.0	...	0	0	0	0	0	2017	12	51	13.179200	0.483000
2017-12-22 12:00:00	0.0	0.0	0.0	1.0	0.0	0.0	0.0	0.0	0.0	0.0	...	0	0	0	0	0	2017	12	51	41.608833	1.361667
...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...
2022-12-21 12:00:00	1.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	...	0	0	0	0	1	2022	12	51	13.609429	0.523857
2022-12-21 15:00:00	1.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	1.0	0.0	...	0	0	0	0	1	2022	12	51	26.405750	0.744250
2022-12-21 18:00:00	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	1.0	0.0	...	0	0	0	0	1	2022	12	51	11.880333	0.199333
2022-12-21 21:00:00	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	...	0	0	0	0	1	2022	12	51	16.568750	0.366250
2022-12-22 00:00:00	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	1.0	0.0	...	0	0	1	0	0	2022	12	51	5.893667	0.513667

	count	mean	std	min	25%	50%	75%	max
Title
#chicagogirl - facebookowa rewolucja	4.0	0.044000	0.032166	0.012	0.02700	0.0380	0.05500	0.088
#jestem m misfit	90.0	0.229278	0.508200	0.000	0.00000	0.0000	0.12425	2.961
#liniaprosta	287.0	0.294787	0.337588	0.000	0.00000	0.1590	0.44350	1.481
#piatkadlarysia	90.0	0.223189	0.305662	0.000	0.00000	0.0745	0.37950	1.449
#tvn pomagajmy sobie	277.0	0.299625	0.602546	0.000	0.00000	0.0080	0.45900	7.156
...	...	...	...	...	...	...	...	...
Zywioly ktore niosa smierc	239.0	0.118828	0.389018	0.000	0.00000	0.0070	0.08650	5.124
Zywnosc 3.0	6.0	0.047500	0.034315	0.010	0.01750	0.0470	0.07200	0.093
Zywnosc jutra	36.0	0.139694	0.163489	0.000	0.00925	0.0565	0.22650	0.574
Zywot michala	7.0	0.203429	0.484094	0.000	0.00000	0.0130	0.05600	1.299
Zza plotu	9.0	0.011889	0.032088	0.000	0.00000	0.0000	0.00000	0.097

	Date_real	Title	Audience	Share	Date_time
0	2022-12-22	Ludzie gor	16.456	1.941	2022-12-22 02:00:00
1	2022-12-22	Ludzie gor	11.909	2.069	2022-12-22 02:45:00
2	2022-12-22	Ludzie gor	0.000	0.000	2022-12-22 03:40:00
3	2022-12-22	Wojny magazynowe w teksasie	0.666	0.151	2022-12-22 04:35:00
4	2022-12-22	Wojny magazynowe nowy jork	0.000	0.000	2022-12-22 05:00:00
...	...	...	...	...	...
2638	2023-02-28	Niewyjasnione tajemnice swiata	10.466	0.340	2023-02-28 23:00:00
2639	2023-02-28	Jak to wyjasnic	19.212	0.902	2023-02-28 23:55:00
2640	2023-03-01	Jak to wyjasnic	24.851	1.534	2023-03-01 00:25:00
2641	2023-03-01	Trojkat bermudzki przeklete wody	16.585	1.452	2023-03-01 00:50:00
2642	2023-03-01	Ludzie gor	16.086	1.926	2023-03-01 01:45:00

	Audience_x	Audience_y
Date_time
2022-12-22 03:00:00	0.959250	2.407802
2022-12-22 06:00:00	9.051500	5.168096
2022-12-22 09:00:00	5.171167	8.140515
2022-12-22 12:00:00	11.316714	15.254249
2022-12-22 15:00:00	11.344250	15.709075
...	...	...
2023-02-21 15:00:00	16.071000	13.756059
2023-02-21 18:00:00	12.263250	9.780672
2023-02-21 21:00:00	21.630000	9.273050
2023-02-22 00:00:00	6.728400	4.028929
2023-02-22 03:00:00	1.308200	2.971485